JMIR Publications

ABSTRACT

Background:

Psychotherapy is a primary treatment for many mental health conditions, yet the dynamic interplay among therapist behaviors, client responses, and the therapeutic relationship remains difficult to disentangle at scale. Traditional process research relies on labor-intensive human coding, limiting the ability to model moment-to-moment therapeutic dynamics across large datasets.

Objective:

This study aimed to develop and validate a computational framework for modeling therapist–client interaction processes and to examine how therapist behaviors and relational qualities shape client disclosure and emotional expression in psychotherapy sessions.

Methods:

We developed automated assessment pipelines using large language models (LLMs) to measure therapist behaviors (empathy, exploration), relational quality (rapport), and client outcomes (self-disclosure, self-directed negative emotion, outward-directed negative emotion). Model-generated scores were validated against human annotations. We then applied these measures to approximately 2,000 hours of psychotherapy transcripts from the Alexander Street corpus. Structural Equation Modeling (SEM) was used to estimate moment-to-moment relationships among therapist behaviors, rapport, and subsequent client responses. Models controlled for prior client state and contextual factors to isolate direct and moderated effects.

Results:

Automated measures demonstrated fair to excellent agreement with human ratings (ICC = 0.45–0.81), with constructs such as rapport (ICC = 0.81) and self-disclosure (ICC = 0.78) falling in the excellent range (ICC = 0.75–1.00), supporting construct validity. SEM analyses revealed that therapist empathy and exploration directly predicted increased client disclosure and shifts in emotional expression. Empathy showed a stronger association with self-directed negative emotions than outward-directed emotions, suggesting increased acknowledgment of internal distress. Exploration was associated with increased disclosure and emotional elaboration. In contrast, rapport did not directly amplify disclosure or emotional intensity; instead, it moderated associations between therapist behaviors and client affect, potentially contributing to reductions in internal emotional distress. Model fit indices supported the adequacy of the structural model (all fit indices within accepted thresholds).

Conclusions:

This study demonstrates that LLM-based measurement combined with structural modeling can capture core therapeutic processes at scale. Empathy and exploration exert direct, moment-to-moment effects on client disclosure and emotional direction, whereas rapport primarily functions as a contextual moderator. These findings provide a computational foundation for precision modeling of psychotherapy processes and offer actionable insights for scalable therapist training and AI-supported clinical education systems.